Prevention of VTE in Nonsurgical PatientsPrevention of VTE in Nonsurgical Patients: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines FREE TO VIEW

Note on Shaded Text: Throughout this guideline, shading is used within the summary of recommendations sections to indicate recommendations that are newly added or have been changed since the publication of Antithrombotic and Thrombolytic Therapy: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Recommendations that remain unchanged are not shaded.

2.3. For acutely ill hospitalized medical patients at increased risk of thrombosis, we recommend anticoagulant thromboprophylaxis with low-molecular-weight heparin [LMWH], low-dose unfractionated heparin (LDUH) bid, LDUH tid, or fondaparinux (Grade 1B).

Remarks: In choosing the specific anticoagulant drug to be used for pharmacoprophylaxis, choices should be based on patient preference, compliance, and ease of administration (eg, daily vs bid vs tid dosing), as well as on local factors affecting acquisition costs (eg, prices of various pharmacologic agents in individual hospital formularies).

2.4. For acutely ill hospitalized medical patients at low risk of thrombosis, we recommend against the use of pharmacologic prophylaxis or mechanical prophylaxis (Grade 1B).

2.7.1. For acutely ill hospitalized medical patients who are bleeding or at high risk for bleeding, we recommend against anticoagulant thromboprophylaxis (Grade 1B).

2.7.2. For acutely ill hospitalized medical patients at increased risk of thrombosis who are bleeding or at high risk for major bleeding, we suggest the optimal use of mechanical thromboprophylaxis with graduated compression stockings (GCS) (Grade 2C) or intermittent pneumatic compression (IPC) (Grade 2C), rather than no mechanical thromboprophylaxis. When bleeding risk decreases, and if VTE risk persists, we suggest that pharmacologic thromboprophylaxis be substituted for mechanical thromboprophylaxis (Grade 2B).

Remarks: Patients who are particularly averse to the potential for skin complications, cost, and need for clinical monitoring of GCS and IPC use are likely to decline mechanical prophylaxis.

2.8. In acutely ill hospitalized medical patients who receive an initial course of thromboprophylaxis, we suggest against extending the duration of thromboprophylaxis beyond the period of patient immobilization or acute hospital stay (Grade 2B).

3.2. In critically ill patients, we suggest against routine ultrasound screening for DVT (Grade 2C).

3.4.3. For critically ill patients, we suggest using LMWH or LDUH thromboprophylaxis over no prophylaxis (Grade 2C).

3.4.4. For critically ill patients who are bleeding, or are at high risk for major bleeding, we suggest mechanical thromboprophylaxis with GCS (Grade 2C) or IPC (Grade 2C) until the bleeding risk decreases, rather than no mechanical thromboprophylaxis. When bleeding risk decreases, we suggest that pharmacologic thromboprophylaxis be substituted for mechanical thromboprophylaxis (Grade 2C).

4.2.1. In outpatients with cancer who have no additional risk factors for VTE, we suggest against routine prophylaxis with LMWH or LDUH (Grade 2B) and recommend against the prophylactic use of vitamin K antagonists (Grade 1B).

Remarks: Additional risk factors for venous thrombosis in cancer outpatients include previous venous thrombosis, immobilization, hormonal therapy, angiogenesis inhibitors, thalidomide, and lenalidomide.

4.2.2. In outpatients with solid tumors who have additional risk factors for VTE and who are at low risk of bleeding, we suggest prophylactic-dose LMWH or LDUH over no prophylaxis (Grade 2B).

Remarks: Additional risk factors for venous thrombosis in cancer outpatients include previous venous thrombosis, immobilization, hormonal therapy, angiogenesis inhibitors, thalidomide, and lenalidomide.

4.4. In outpatients with cancer and indwelling central venous catheters, we suggest against routine prophylaxis with LMWH or LDUH (Grade 2B) and suggest against the prophylactic use of vitamin K antagonists (Grade 2C).

5.1. In chronically immobilized persons residing at home or at a nursing home, we suggest against the routine use of thromboprophylaxis (Grade 2C).

6.1.1. For long-distance travelers at increased risk of VTE (including previous VTE, recent surgery or trauma, active malignancy, pregnancy, estrogen use, advanced age, limited mobility, severe obesity, or known thrombophilic disorder), we suggest frequent ambulation, calf muscle exercise, or sitting in an aisle seat if feasible (Grade 2C).

6.1.2. For long-distance travelers at increased risk of VTE (including previous VTE, recent surgery or trauma, active malignancy, pregnancy, estrogen use, advanced age, limited mobility, severe obesity, or known thrombophilic disorder), we suggest use of properly fitted, below-knee GCS providing 15 to 30 mm Hg of pressure at the ankle during travel (Grade 2C). For all other long-distance travelers, we suggest against the use of GCS (Grade 2C).

6.1.3. For long-distance travelers, we suggest against the use of aspirin or anticoagulants to prevent VTE (Grade 2C).

7.1. In persons with asymptomatic thrombophilia (ie, without a previous history of VTE), we recommend against the long-term daily use of mechanical or pharmacologic thromboprophylaxis to prevent VTE (Grade 1C).

This article focuses on prevention of VTE in nonsurgical populations. Because they are addressed in other chapters in these guidelines,^1,2 we do not include prevention of VTE in patients with trauma and spinal cord injury and in patients with ischemic and hemorrhagic stroke.

Adverse consequences of unprevented VTE include symptomatic DVT and pulmonary embolism (PE), fatal PE, chronic postthrombotic syndrome, and increased risk of recurrent VTE. We consider the desirable and undesirable consequences of antithrombotic prophylaxis to prevent VTE in the following populations/patient groups: (1) hospitalized acutely ill medical patients, (2) critically ill patients, (3) patients with cancer receiving cancer treatment in the outpatient setting, (4) patients with cancer with indwelling central venous catheters (CVCs), (5) Chronically immobilized patients, (6) long-distance travelers, and (7) asymptomatic persons with thrombophilia. We also consider the use of statins (HMG-CoA reductase inhibitors) to prevent VTE. Table 1 describes the question definition (population, intervention, comparator, and outcome) and eligibility criteria for studies considered in each section of this article.

The methodology of these guidelines follows the general approach of Methodology for the Development of Antithrombotic Therapy and Prevention of Thrombosis Guidelines. Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines in this supplement.³ In brief, panel members conducted literature searches to update the existing evidence base, seeking systematic reviews and trials published since the previous iteration of the guidelines, and rated the quality of the evidence using the Grading of Recommendations Assessment, Development, and Evaluation framework. The panel considered the balance of benefits and harm, patients’ values and preferences, and patients’ context and resources to develop weak or strong recommendations. In this article, we identified three areas with sparse high-quality evidence: (1) the benefits of prophylaxis as measured by reduction of the incidence of symptomatic VTE events, (2) resource use and cost-effectiveness, and (3) the benefits of screening strategies for VTE in nonsurgical patients.

We selected similar patient-important outcomes across recommendations. These include symptomatic DVT, PE, death from PE, major bleeding, heparin-induced thrombocytopenia (HIT), and mechanical thromboprophylaxis complications (when applicable). In addition, for patients with CVCs, we include catheter failure as an outcome.

As the mortal outcome of greatest interest, when data were available, we have chosen treatment-related mortality (PE deaths, hemorrhagic deaths). For pharmacologic interventions, when available, we provide data on fatal bleeding and fatal intracranial bleeding as a subset of all-cause mortality, and for the outcome of major bleeding, when available, we provide data on intracranial bleeding and GI bleeding (the most common type of “critical organ” bleeding expected in nonsurgical populations). Given that anticoagulants used to prevent VTE are administered for short periods of time, major bleeding and fatal bleeding are likely to be rare events, except during critical illness.

Little is known about the distribution of patients’ values and preferences in the context of VTE prevention in nonsurgical settings. In developing the recommendations for this guideline, panelists made estimates of patients’ values and preferences often using indirect data from other settings (eg, values and preferences that pertain to anticoagulation in atrial fibrillation).

In our populations, the weights (relative importance) given to the harmful effects (disutilities) of the most representative types of critical organ bleeding, namely GI or, less commonly, intracranial bleeding, will greatly impact the tradeoff between desirable and undesirable consequences of antithrombotic therapy. There are limited data to guide us with respect to the relative impact of VTE events vs bleeding events on patient-perceived state of health; available evidence suggests values and preferences for treatments and for health states vary appreciably between individuals.⁴

In a values rating exercise, Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines panelists used a “feeling thermometer” with anchors at 0 (representing death) and 100 (representing full health) to rate patient scenarios for various clinical outcomes in terms of the value placed on a year in which the events depicted in the scenario occurred.³ Median ratings were similar for the outcomes of symptomatic DVT, PE, and catheter thrombosis (80, 75, and 80, respectively) and severe GI bleeding (75), whereas the median rating for intracranial bleeding (stroke scenario) was 40. Therefore, we used 1:1 ratio of symptomatic VTE to major extracranial bleeding and 2.5:1 ratio of symptomatic VTE to intracranial bleeding for tradeoffs.

We considered that preventative and screening recommendations require higher-quality evidence supporting benefit than therapy recommendations. This decision is a value-based judgment. In making our recommendations, when there is uncertain benefit and an appreciable probability of important harm or patient burden associated with treatment, we recommend against such treatments.

In making clinical recommendations, guideline developers need to consider the balance of benefits and harms in terms of absolute treatment effect on patient-important symptomatic events in addition to relative measures of risk. The panelists of the four articles dealing with VTE prevention faced challenges in finding these data and developed several possible approaches for estimating the effect of prophylaxis on the incidence of symptomatic VTE events. In this article, we used two different approaches for hospitalized patients in non-critical care settings and for critically ill patients, based on the availability of data.

Since medical patients have a significantly heterogeneous risk for VTE, the guideline panel sought to evaluate preventive strategies in two different strata of patients (low risk and high risk). We decided against simply using as the baseline estimate the pooled average risk of DVT (0.8%) and PE (0.4%) reported in the control arms of the randomized controlled trials (RCTs) of thromboprophylaxis in hospitalized medical patients, as it is evident from the trials’ eligibility criteria that patients with heterogeneous risk were enrolled (Table S1) (Tables that contain an “S” before the number denote supplementary tables not contained in the body of the article and available online. See the “Acknowledgments” for more information.). Also, there is uncertainty about the generalizability of trial results to other populations, as in many of the trials the ratio of patients screened to patients enrolled was very high (eg, ≥ 100), and probable underestimation of absolute numbers of symptomatic events, as patients diagnosed with asymptomatic DVT via trial-mandated screening tests are typically treated with anticoagulants. Incidence estimates from most observational studies were unsatisfactory because they were not stratified by the use of thromboprophylaxis and were also reported in very heterogenous populations (Table S2).

To estimate baseline risk for patients with low and high VTE risk, we used data from risk assessment models (RAMs). Several RAMs have been proposed for use in hospitalized medical patients (Table S3).^5‐7 Limitations of most RAMs include lack of prospective validation, applicability only to high-risk subgroups, inadequate follow-up time, and excessive complexity.

In a prospective observational study of 1,180 inpatients, a predefined RAM (Padua Prediction Score, modified after Kucher⁸) assigned points to 11 common VTE risk factors (Table 2)⁹ and categorized hospitalized medical patients as low risk (< 4 points; 60.3% of patients) or high risk (≥ 4 points; 39.7% of patients) for VTE. Attending physicians were not notified of their patients’ risk categories. Patients were followed for symptomatic VTE for 90 days. VTE occurred in 11.0% of high-risk patients who did not receive prophylaxis vs 0.3% of low-risk patients, a > 30-fold difference in risk (hazard ratio [HR], 32.0; 95% CI, 4.1-251.0). Among 711 low-risk patients, two (0.3%) developed VTE (1 PE, 1 PE with DVT). Among 283 high-risk patients who did not receive prophylaxis, the risk of DVT was 6.7%, nonfatal PE 3.9%, and fatal PE 0.4%. Hence, for baseline risk for low- and high-risk strata, we used risk estimates provided by the Padua Prediction Score.⁹ Despite the limitations of this risk model (small number of events, suboptimal validation), this model provides the best available basis for judging hospitalized patients’ risk.

We considered a number of options for baseline risk of major bleeding. We considered bleeding events reported in the Padua prediction score study. However, this study stratified bleeding events according to thrombosis risk, not bleeding risk (1 of 283 in the low VTE risk group [0.4%; 95% CI, 0.0-2.0] and 1 of 711 in the high VTE risk group [0.1%; 95% CI, 0.0-0.8]).⁹ We also considered bleeding events in a large observational study by Decousus¹⁰; however, this study did not report bleeding according to use of pharmacoprophylaxis. Therefore, we chose to use 0.4% (19 of 4,304) derived from the control arm of trials of thromboprophylaxis in medical patients as the estimate of baseline risk of major bleeding (section 2.1). Where possible, we presented data on intracranial bleeding separately from major bleeding events.

Critical care trials have routinely screened patients for asymptomatic DVT, which are usually promptly treated if detected. Hence, an accurate estimate of risk of symptomatic DVT is not available from trials of critically ill patients receiving no prophylaxis, and PE events are generally rare. We used two approaches to estimate the baseline risk and absolute risk difference in critically ill patients. When symptomatic events were reported, such as DVT in the trials by Shorr et al¹¹ and in PROTECT,¹² we used these data directly to estimate the baseline risk, relative risk (RR), and risk difference. When symptomatic events were not reported in the trials, such as the PE outcome in trials that compared unfractionated heparin (UFH) or low-molecular-weight heparin (LMWH) vs placebo, we opted to use a baseline risk derived from symptomatic PEs reported in three observational studies.^13‐15 The risk ratio (RR) was derived from the trials in which events were likely a mix of symptomatic and asymptomatic events. The former approach has the advantage of directness but may suffer from imprecision and poor applicability. The latter approach requires imputations that make the evidence indirect.

Hospitalization for acute medical illness is associated with an eightfold increased risk of VTE¹⁶ and accounts for about one-fourth of all VTE events in the community.^17,18 Among hospitalized patients, 50% to 75% of VTE events, including fatal PE, occur in those hospitalized on the medical service.^16,19 Risk factors for VTE in hospitalized medical patients include intrinsic factors, such as increasing age (especially > 70 years), previous VTE, known thrombophilia, and various comorbid illnesses, such as cancer, heart failure, or respiratory failure, and extrinsic factors, such as immobilization for ≥ 3 days and hormonal medications^5,20‐22 (Table 2).⁹

A recent multinational observational study reported on risk factors at admission that were independently predictive of in-hospital bleeding (the analysis combined major and nonmajor clinically relevant bleeding) among 10,866 hospitalized medical patients. The strongest risk factors were active gastroduodenal ulcer, bleeding in 3 months before admission, and platelet count < 50 × 10⁹/L, followed by age > 85 years, hepatic failure, severe renal failure, and ICU or critical care unit admission (Table 3).¹⁰ Although data on incidence of bleeding were not provided separately by use vs nonuse of prophylaxis (overall rate of major bleeding was 0.76%), the above variables remained predictive of bleeding when the model was adjusted for pharmacologic prophylaxis. A bleeding risk score that included these and additional variables was developed by the authors, who reported that more than one-half of all major bleeding episodes occurred in the 10% of hospitalized medical patients who had a bleeding risk score ≥ 7.0.

Although this risk score is complex and has not yet been validated, the panel considered patients to have an excessive risk of bleeding if they had multiple risk factors or had one of the three risk factors with the strongest association with bleeding (OR > 3.0): active gastroduodenal ulcer, bleeding in 3 months before admission, and platelet count < 50 × 10⁹/L.

We used data from three contemporary, high-quality systematic reviews to assess the benefits and harms of anticoagulant prophylaxis vs no prophylaxis in hospitalized, acutely ill medical patients.^23‐25 In general, the trials included acutely ill hospitalized patients (typically, the mean age of enrolled patients was > 65 years) admitted for congestive heart failure, severe respiratory disease, or acute infectious, rheumatic, or inflammatory conditions, who were immobilized and had one or more additional VTE risk factors including but not limited to age > 40 years, active cancer, previous VTE, or serious infection (Table S2). Prophylactic anticoagulant regimens included low-dose unfractionated heparin (LDUH) tid, LDUH bid, various LMWHs, and fondaparinux. Duration of use of prophylaxis ranged from 6-21 days or discharge from hospital, whichever came first. In all trials, routine screening for DVT was performed.

Meta-analysis of these trials demonstrates that anticoagulant thromboprophylaxis is associated with significant reduction in fatal PEs (RR, 0.41; 95% CI, 0.22-0.76; two fewer per 1,000 [95% CI, from one fewer to three fewer]). When we apply the relative effect of anticoagulant thromboprophylaxis obtained from these meta-analyses to baseline risks obtained from risk assessment models, we find that thromboprophylaxis is associated with a reduction in symptomatic DVT (RR, 0.47; 95% CI, 0.22-1; one fewer per 1,000 [95% CI, from one fewer to 0 fewer] in low-risk patients; 34 fewer per 1,000 [95% CI, from 51 fewer to 0 fewer] in high-risk patients). The effect on nonfatal PE, major bleeding, and all-cause mortality was not statistically significant and is described in terms of relative and absolute effects (Table 4, Table S4). No trial reported the incidence of HIT.

Based on these data, the panel judged that moderate-quality evidence suggests that thromboprophylaxis is effective in reducing symptomatic DVT and fatal PE in acutely ill, hospitalized, immobilized medical patients who have characteristics similar to those enrolled in the above RCTs, and moderate-quality evidence suggests a modest relative and very small absolute increase in bleeding risk. Based on the above RCTs, the panel considered that providing prophylaxis for 6 to 21 days, until full mobility is restored or until discharge from hospital, whichever comes first, is a reasonable approach. The recommendation to prophylax applies only to the higher-risk patients (Table 2). In low-risk patients, VTE is too infrequent to warrant prophylaxis.

LDUH and LMWH (enoxaparin, nadroparin, or certoparin) have been directly compared in five RCTs.^26‐30 Eligibility criteria for RCTs of LDUH vs LMWH in hospitalized medical patients were similar to trials of any anticoagulant vs none to prevent VTE and are shown in Table S5. In all trials, routine screening for DVT was performed. Dosing of LDUH was tid in four trials and bid in one trial.²⁶

Pooled results failed to exclude benefit or harm for LMWH vs LDUH for the outcomes DVT (RR, 0.77; 95% CI, 0.50-1.19), PE (RR, 1.00; 95% CI, 0.28-3.59), overall mortality (RR, 0.89; 95% CI, 0.65-1.23), and HIT (RR, 0.50; 95% CI, 0.05-5.48) (Table 5, Table S6). Pooled results for major bleeding suggest a large relative protective effect of LMWH (RR, 0.48; 95% CI, 0.24-0.99) and small absolute (five fewer; 95% CI, 0-7 fewer) reduction in bleeding events per 1,000 patients treated. Evidence is consistent with a similar effect of LMWH and UFH on reduction in thrombosis in acutely ill hospitalized medical patients (though imprecision is such that effects could, in relative terms, be appreciably greater in one treatment or the other). The potential for less bleeding with LMWH represents a benefit that is small, and it may be very small.

The International Medical Prevention Registry on Venous Thromboembolism (IMPROVE), a registry of 15,156 acutely ill hospitalized medical patients enrolled at 52 hospitals in 12 countries, documented marked variation in practices in dosing frequency of LDUH used to prevent VTE. LDUH was prescribed tid in 54% of patients from the United States compared with bid in 85% of non-US patients.³¹

There have been no head-to-head trials comparing bid vs tid LDUH to prevent VTE in hospitalized medical patients. We conducted a mixed-treatment comparison meta-analysis of 16 RCTs that enrolled hospitalized nonsurgical patients at risk for VTE and compared LDUH bid, LDUH tid, or LMWH to each other or to an inactive control.³² The RR and 95% credible intervals comparing LDUH tid to LDUH bid for DVT, PE, death, and major bleeding (all were indirect comparisons) were 1.56 (0.64-4.33), 1.67 (0.49-208.09), 1.17 (0.72-1.95), and 0.89 (0.08-7.05), respectively. Due to a lack of reporting, we could not perform this analysis for the outcome HIT. The low-quality evidence from these indirect comparisons provides no compelling evidence that LDUH tid dosing, compared with bid dosing, reduces VTE or causes more bleeding. A future randomized trial comparing these agents is unlikely, considering the large sample size that would be required to demonstrate a significant difference, which, if it exists, is undoubtedly small. From a patient preference perspective, twice daily injections are likely to be preferred and better tolerated than thrice daily injections.

Almost all cost-effectiveness analyses in this population have reported costs per VTE or death averted with the use of anticoagulant prophylaxis, but few studies have reported costs per quality-adjusted life-year gained to compare against preexisting benchmarks. Two studies that reported incremental costs of $65 to $2,534 per quality-adjusted life-year gained over no prophylaxis were both sponsored by the pharmaceutical industry.^33,34 In populations at sufficiently high risk (Tables 2, 6, Table S7), pharmacoprophylaxis is likely to be favorable from a resource standpoint for preventing VTE.^35,36 The comparison between different types of prophylaxis, however, is less clear.

Several studies have suggested that choosing LMWH over LDUH is cost neutral, or even cost saving.^37‐41 However, the quality of these analyses is moderate at best. First, many of the authors have had financial disclosures with the pharmaceutical industry, and whether these ties influence the cost-neutral or cost-saving results of LMWH over LDUH is unclear. Second, the performance estimates used in most of these studies have been extracted from the Medical Patients with Enoxaparin (MEDENOX) trial, which did not directly compare LMWH to LDUH